Super Intelligence (SI) is the term the US government adopted on September 29, 2026 for the generative tools creators have been calling AI, and the earlier guides in this track covered those instruments — words with an SI writing partner, SI images, motion and sound, and the rules around them. This one covers the orchestra: how the pieces assemble into a working practice, and what changes about being a creator when generation itself becomes nearly free. The short version: every stage of the creative process gets faster except one — and that one becomes your entire profession.
Map SI to stages, not to projects
The unproductive question is 'should I use SI for this project.' The productive one is 'which *stage* of this project does SI serve' — because the answer differs by stage far more than by project:
- Ideation: wide open. Volume is the whole point here, and volume is what generation does. Fifty title options, twenty visual directions, a dozen structural approaches — generated exploration costs minutes, and bad options die free. The discipline is generating *divergently* (ask for the safe version, the weird version, the wrong-on-purpose version) rather than accepting the first competent cluster.
- Drafting and production: selectively open. This is where the medium-specific crafts from the earlier guides live — SI drafts and elements as raw material, with the load-bearing parts kept human. The ratio varies honestly by medium and stakes; the constant is that something of yours is in the substrate.
- Finishing: mostly human. The last ten percent — the edit that makes it cohere, the cut that makes it land, the detail-checking that each medium's known failure points demand — resists delegation, and it is precisely the ten percent audiences and clients are paying for.
- The idea of the idea: entirely human. What is worth making, for whom, and why now. No stage of any pipeline supplies this, and every other stage amplifies whatever answer you bring to it.
Taste is the bottleneck, and the bottleneck is the job
When generating a hundred options costs nothing, choosing among them becomes the scarce activity. This is the deepest change SI brings to creative work: the profession migrates from making to selecting-and-shaping — from craftsman toward editor, director, curator. Two practical consequences follow. First, taste needs *calibration inputs*: the creators thriving in this mode consume more reference, study more work outside their genre, and articulate more precisely *why* option seven beats option twelve — because articulated taste can steer the next generation round, and inarticulate taste can only reroll. Second, taste needs *protecting from fatigue*: reviewing a hundred near-misses dulls judgment in a way reviewing five sketches never did. Short selection sessions, culling hard and early, and a 'kill file' of rejected directions are the mundane defenses.
Volume is a strategy decision, not a default
The economics whisper constantly: you could ship ten times more. Sometimes that whisper is right — content operations, client volume work, and experimentation all genuinely benefit from scale. But volume and identity trade against each other at the margin, and the feeds already show the failure mode: creators who let generation set their pace, publishing at machine speed with human judgment spread ever thinner, converge on the same median output the tools default to — and audiences, whose filtering instincts are sharpening fast, scroll past. The durable position is the deliberate one: use the speed to *raise the floor and buy back time* — more drafts explored, more finishing on what ships, fewer things shipped better — rather than to flood. Scarcity of attention did not go away; it moved to the audience's side of the screen.
The SI-era practice, in habits
What this looks like as a week rather than a philosophy: an exploration session where generation runs wide and cheap; a selection session, fresh-eyed and ruthless; production where SI supplies material and you supply the spine; a finishing pass that no deadline is allowed to eat; and the process records kept throughout. Plus one meta-habit this field uniquely demands — a small, regular time-box for evaluating new tools, because the capabilities genuinely shift every few months, and both the creator who chases every release and the one who evaluated once in 2024 are getting it wrong. The instruments will keep improving. The orchestra is yours.